Outer Concrete Containments of LNG-Tanks Design against Thermal

Shock
Josef Roetzer
Dywidag International, Munich, Germany

Theodor Baumann
Munich, Germany

ABSTRACT: The stability and tightness of the concrete outer tank has to be guaranteed by an appropriate design also under cryogenic conditions. The relevant codes provide a lot of formal regulations, but give no precise indications for analytical procedures and criteria which have to be applied in the design. Hence, the following paper deals with the behaviour of reinforced and prestressed concrete sections in direct contact with
LNG, considering thermal strains and consecutive crack formation. Mechanical models, which have to be
clear and simple, are discussed. The course of sectional forces and displacements due to a temperature gradient of 180C after failure of the inner tank are outlined for areas below and above the LNG level. Substantial
design criteria are proposed and discussed. Essential in this respect are the thickness of the residual compressive zone, the reinforcement steel stresses and the characteristic crack width. By means of ingenious models, a
way for the direct understanding of the coherence between the strains and deformations imposed by the temperature gradient and the above design criteria is pointed out.
1 FUNCTION OF LNG-TANKS
The storage of liquefied natural gas (LNG) at -165C
is an economic significant and technical demanding
task. The concrete outer tank protects the sensitive
inner steel tank against external hazards and serves
as catch basin in case of failure of the inner tank.
This paper deals with the behaviour of the concrete
outer tank under the extreme temperature loading
caused by the direct contact with LNG.
The application of LNG-technology is based on
the physical behaviour of LNG. By cooling down to
-165C the gas is liquefied with a consecutive reduction in volume of 1/600. Thereby the transportation
with ships and the appertaining storage in tanks become economically feasible. A state-of-the-art design requires a prestressed concrete outer tank which
is able to operate as a catch basin in case of liquid
spill. This conception requires even in the emergency case a closed container, which is able to prevent the uncontrolled escape of gas clouds into the
environment.
Containments according to this design philosophy
are called full containment tanks. They consist of a
concrete outer tank and a steel inner tank. The concrete outer tank is composed of a bottom slab, a
prestressed wall shell, often with a ring beam on top
of the wall, and a reinforced concrete roof. The hydrostatic LNG pressure requires prestressing of the
wall in hoop (horizontal) direction. The size of the

vertical prestressing depends primarily on the internal gas overpressure. The liquid gas is stored in the
steel inner tank. Adequate ductility of steel at -165C
requires a minimum nickel content of 9%. The inner
steel tank is open on top. The LNG is pumped into
and out of the tank by cryogenic pumps suspended
on the concrete roof.

Figure 1: Typical components of a LNG tank

Heat supply is limited by an all side insulation to

reduce the evaporation rate of the LNG, which is at
-165C close to the boiling point. Hence, between

concrete outer tank and steel inner tank, a thermal

insulation layer of about 1m is necessary.
For layout and design of the concrete outer tank
an appropriate assessment of the effect of extreme
temperatures is essential. The design criteria are impermeability and stability. Even some throughcracks indicate no catastrophic scenario, as long as
the crack width is limited. LNG which is infiltrating
into the cracks passes inside the wall into the gaseous state and reaches the wall outer surface not as
fluid. The design of the reinforced concrete structure
requires the consideration of the nonlinear material
behaviour and crack formation. Thereby the sectional forces due to constraint are reduced to a fraction of the values of the uncracked state. The structural analysis shall demonstrate that the cracked wall
section is able to bear the pressure from leakage and
internal gas overpressure. The common approach of
separate assessments for load and for restraint of
thermal deformation makes no sense. The analysis of
strains is only possible for the common action of
loads and imposed deformations.
For design of the reinforced and prestressed concrete walls the liquid spill is the dominant load case,
which is furthermore highly complex. For the accuracy to be reached, the understanding of the behaviour of the cracked structure is the most important
item. Therefore this paper deals with the relation between the temperature gradient and the resulting
stresses, strains and crack widths. Substantial design
criteria, which are able to ensure the impermeability
of the concrete wall, are discussed and proposed.
2 CODES AND THERMAL PROCESS
CONDITIOS
The relevant codes e.g. BS 7777 or EN 14620
provide a lot of formal regulations but give no precise indications for analytical procedures and criteria
to be applied in the design. The common way of an
detailed FEM-calculation with highly sophisticated
algorithms and nonlinear material behaviour makes
no sense, if the physical background is not understood by the user.
BS 7777 requires in case of liquid spill the verification of an average concrete stress of 1 N/mm2 in
the residual compressive zone and that the compressive zone is large enough to ensure impermeability.
Other codes additionally define the minimum compressive zone to 80mm respectively 100mm and
10% of the cross section height.
For a better understanding of the load case liquid
spill, the time dependent events, leakage and cool
down of the concrete wall are examined separately.
The inner tank has an usable volume of 160,000m3
and the usable volume on the perlite wall insulation
is about 4000m3. If the leakage is small and the dis-

charge amounts some liters per second, it takes a

couple of days until a balanced level is reached.
The cooling down process of the entire wall section is finished, when the temperature gradient
within the wall is nearly linear (steady state). For an
80cm concrete wall this process takes about three
days (Roetzer et al. 2006). Following this final stage
is examined.
3 PROOF OF THE TANK WALL AGAINST
DIRECT EFFECT OF LNG
3.1 General
In case of failure of the inner tank, LNG runs into
the space filled with perlite between inner and outer
tank and gets in direct in contact with the concrete
wall. The inner concrete surface suddenly cools
down from +35C to -165C. From this results an effective steady state temperature gradient TL of
180K. Concrete and reinforcement at the layer want
to shorten with TL = TTL, but the warmer inner
wall areas prevent this. Hence, narrow distributed
cracks at the concrete surface are generated. The reinforcement stress reaches values in the range of (s
= EsTTL 200,00010-5180 360MPa). The
subsequent cooling of the wall reduces the reinforcement stress but increases the crack depth. The
steady state temperature gradient represents the most
unfavorable design case for stability and impermeability of the catch basin. Temperature caused deformations and stresses are much larger than those
generated from hydrostatic pressure. For design and
analysis it is essential to take the crack formation
into account.

Figure 2: Typical dimensions and hoop forces of a LNG tank

The practical effects are demonstrated for a

160,000m3 tank with a wall thickness of 0.80m. The
hoop prestressing has to be set higher than the forces
resulting from LNG filling. The residual compres-

sion stress is taken in this example to (nL+nph)/h =

6.5 8.5 = -2.0MPa, see Figure 2.
Figure 3a shows a top view of a wall section
without any restraint of deformation (length 2l). The
temperature gradient TL and the related expansion
gradient TL = TTL are generating a shortening
l = 0.5TLl and a corresponding rotation of =
TLl/h at the end of the section. Actually the sections of a cylinder shell are able to shorten, but are
not able to distort.

prestressing in hoop direction or from dead load, internal gas overpressure and vertical prestressing in
meridian direction. In the idealisations of Figure 3, n
= 0 has been assumed. In the real design, a residual
compressive force n < 0 should be verified.
Temperature related constraint forces are generated not only due to restraint of the gradient TL,
but also due to restraint of center line shortening (TII
L = c). The required continuous deformation transition from cylinder wall to the not cooled down bottom slab generates in the lower wall region
additional bending moments (edge disturbances),
whose appearance is explained in Figure 7. This
topic is handled in the section 3.3.
3.2 Restraint of the imposed curvature TL / h
The thermal shortening of the inner side in direct
contact with LNG is assumed to TL = TTL
1.8. For the superposition of the consecutive curvature TL/h with the liquid pressure (load case liquid spill) the following values are required:
xc thickness of the residual compressive zone
s1 inner reinforcement steel stress at the crack
wc characteristic crack width at inner wall face
They can be obtained directly with the relations
summarized in Figure 4. The detour of a calculation
of sectional forces from loading and restraint regarding cracked cross sections with a subsequent dimensioning is not required. The direct approach allows a
better understanding of the mechanical significance
of the assumptions to be defined.
As a result of the distinct cracking, the tension
stiffening effect is low. The ratio k = s1/s1m (maximum to average steel stress) is set to 1.20. The crack
widths can be calculated acc. to DIN 1045-1 or acc.
to the more consistent continuous crack theory of
Noakowski (Noakowski & Schfer 2003).

Figure 3: Deformation and cracking of a concrete wall due to

direct contact with LNG

Therefore, in hoop direction as well as in meridian direction a restraint moment mI is generated (see
Fig. 3b). Distinct cracks are generated, as the appertaining tensile force cI is multiple to the tensile
strength of concrete. Through-cracks appear if no or
only slight reinforcement is placed (Fig. 3c). In case
of higher reinforcement content, a restraint moment
mII according Figure 3d is generated. The shortening
of the neutral axis is calculated with l = xc /hTLl
= cl.
The dependence of the strain conditions which
are appertaining to the restraint of TL from the
relevant parameters are investigated in section 3.2.
The longitudinal force n results either from liquid
pressure, internal gas overpressure and horizontal

Figure 4: Sectional strains and forces for LNG-spill

The prestressing forces nph in horizontal (hoop)

and npv in vertical direction, respectively, are included in the total normal force nh and nv. The
prestressing tendons are placed in the outer third of

the wall, i.e. near to the neutral axis of the cracked

cross section. Therefore the influence of concrete
cracking on the tendon forces can be neglected. The
eccentricity of the prestressing tendons generates no
bending moments, because the shape of the cylinder
shell prevents the appertaining curvature.

considered dimensions, an increasing wall thickness

at bottom slab (haunch) cannot be dispensed with.
Reasons and consequences of the edge disturbance at the lower ending of a cylindrical shell can
be understood easier, if in a first step the deformations are treated without a connection of shell with
bottom slab. Figure 7 shows some typical deformation patterns resulting from the thermal curvature
without and with restraint. If the wall could be deformed without restrain, due to TL = 1.8 an arch
shaped bending of 325mm would occur in vertical
direction. In reality this deformation is prevented by
the restraining stresses as shown in general in Figure
3d and in detail in Figure 6. A residual compressive
stress of n/h = -2MPa is related to a compressive
strain c = -0.53, which represents also the shortening of the neutral axis TLII in hoop direction. The
radius of 40m is thereby shortened by 21mm. The
shape of the continuous transition to the bottom slab
causes vertical moments and hoop forces, which are
designated as edge disturbance.

Figure 5: Steel stress, residual compressive zone, crack width

and stiffness in dependency from the ratio n/h

Figure 5 and 6 apply to the tank shown in Figure

2. The wall thickness is set to 0.80 m and the reinforcement to ds = 25mm/150mm. This corresponds
with a reinforcement ratio of s1 = s2 = 0.4%. The
dependency from n/h demonstrates the effect of the
prestressing level on the governing criteria of cryogenic design. According to Figure 5 and table 1, the
characteristic crack width does not exceed a value of
0.50mm even for a reduction of the residual compressive force (nh or nv) to zero.
Table 1: Effect of the residual compressive stress
Figure 6: Typical states of strain for restraint curvature

On the basis of the assumptions defined in Figure

4, the above data are exact results for areas outside
of discontinuity zones, such as the liquid level and
the bottom edge disturbance. The results are suitable
also for the necessary check and verification of programs for a more complex computational analysis.
3.3 Edge disturbance by restraint of deformation
Following the cylindrical wall is considered at the
transition to the bottom slab. Thereby a constant wall
thickness is assumed in order to demonstrate clearer
the principal influences. Because of the same reason
a corner protection is not included in this analysis.
However in the practical design, for tanks of the

The curve of Figure 8b is based on theory of thin

shells of a not cracked concrete section. The curve is
also valid for cracked sections, if the ratio from extension stiffness to flexural stiffness (DI/BI = DII/BII)
does not change. Figure 5 confirms an approximately
constant ratio. Therefore the sectional forces of Figure 8c and 8d are derived by multiplying elongation
and curvature acc Figure 8b with the stiffness DII respectively BII.
A typical ratio DII/DI can be derived easily by
means of Figure 6. A reduction of the resulting compressive stress from -8MPa to -2MPa respectively an
increase of the hoop tension force from 6.4MN/m to
1.6MN/m causes an increase of strain L = 0.4.
This corresponds to a cracked extension stiffness of
DII = 4.8/0.4 = 12,000 MN/m. In a not cracked
state, the stiffness is calculated with DI = Ech =
34,5000.8 = 27,600 MN/m. Therefore the ratio

DII/DI is resulting to 12,000/27,600 = 0.44. This is in

good correlation with the graph in Figure 5.
A hoop tension force nh greater than 1.6 MN/m
respectively n/h 2 MPa is occurring up to a height
of 4m above the bottom slab as shown in Figure 8c,
and vertical through-cracks would occur with this
design. This is one of the reasons for the necessity of
a thermal corner protection.

Figure 7: Displacement in radial direction

The curvature 1/R2 shown in Figure 8b generates

in meridian direction tensile stresses in vertical direction at the inner face (Fig. 8d). They superpose
with the gradient TL in the same direction (Fig.
8e). In contrast to that, the maximum curvature 1/R1
at the wall junction acts in contrary direction.
Figure 9 and table 2 show strains and other relevant design data for the action of TL only (case A
and B) and for the superposition of TL with 1/R2

(case C and D). The resulting vertical stress nv/h is

assumed to -2MPA (case A and C) and 0MPa (case
B and D). The non prestressed reinforcement is assumed to ds = 25mm/150mm. Dead load of wall and
roof cause a compression stress of nv/h -1MPa.
The size of the required vertical prestressing depends
on the internal gas overpressure. The variation of
nv/h between 0 and -2MPa covers the range to be
expected in practice.
By increasing the residual compressive stress nv/h
from 0 to -2MPa, the steel stress is reduced from
395MPa to 344MPa and the crack width from about
0.6mm to about 0.5mm at the location of R2 acc.
Figure 8. In the area above the edge disturbance, a
reduction from 324 to 276MPa and from about
0.5mm to about 0.4mm can be achieved.
Crack widths of up to 0.5mm seem acceptable in
the special situation of LNG-spill. Because of its
short duration they cause no problem with regard to
corrosion of reinforcement.
The main aspect is the reparability and structural
integrity of the outer containment after such a hazard. This is not endangered by short duration cracks
of up to 0.5mm. As for the high stresses in the reinforcement at the inner face, the only reasonable way
consists in ensuring a sufficient ductility under the
given cryogenic conditions rather than increasing
prestressing or reinforcement. The reinforcement
bars must sustain the imposed strain of 1.3 to 1.9
without risk of rupture. With regard to a safety concept, the ductility should be guaranteed for even
higher strains. The conventional notch impact tests
(e.g. on charpy specimens) seem not sufficient for
the verification of the required ductility considering
the real type of notch stresses at the ribs caused by
the high bond stresses at cracks.

Figure 8: Edge disturbance for typical dimensions (continuous wall thickness 0.80m, no haunch, no Thermal Corner Protection)

Table 2: Values with additional edge disturbances

tom slab and roof ring beam and other discontinuities make sense.

Figure 10: Course of bending moment for different LNG levels

5 CONCLUSION

Figure 9: Relevant strains and stresses in vertical direction

4 PRACTICAL DESIGN OF OUTER

CONCRETE CONTAINMENTS
Edge disturbance and discontinuity zones with regard to haunches, thermal corner protection, variable
liquid levels and other influences can be handled
only by computational analysis with suitable programs. Thereby all the questions and assumptions
discussed in the previous sections have to be dealt
with. First of all it has to be verified, that the applied
program is able to model realistically the liquid
loads and the imposed thermal deformations, strains,
stresses and forces in the cracked structure. This requires thorough tests of the algorithms, which have
to depict correctly the basic behavior outlined in the
above examples. It would be a fundamental error to
expect that even high level standard programs generate correct results automatically only by increasing
the accuracy of the element net and the iteration
process.
Figure 10 shows the results of an analysis (Roetzer et al. 2006) on the simplified system without
haunch and thermal corner protection. The moments
given for the highest liquid level have to be compared with the superimposed moments of Figure 8d.
The correspondence is satisfying.
Only if this first step of verification has been
completed successfully further analyses of more
complicated systems with edge disturbances at bot-

The outer concrete tank operates as collecting basin,

which is exposed to LNG-pressure and thermal
shock only in hazard scenarios. Detailed considerations of the problems connected with such a catastrophic event are not always taken seriously, maybe
because of the low occurrence probability and because of the opinion that the real course of damage is
neither predictable nor afterwards reconstructable.
The corresponding codes and standards doe not
really refute this interpretation. They provide a lot of
formal regulations, but give no precise indications to
the basic requirements. Such an approach is not justifiable, neither economically nor in the face of the
demanded protection of health and safety.
The effort for a consistent analysis and design of
the components and assemblies and for the development and procurement of appropriate materials
with regard to a safe and reliable behaviour also in
hazard situations is minor in comparison to possible
damages.

6 REFERENCES:
BS 7777. 1993. Flat Bottomed, Vertical, Cylindrical Storage
Tanks for Low Temperature Service. Part 3: Recommendations for the Design and Construction of Prestressed and
Reinforced Concrete Tanks. London: BSI
DIN 1045-1. 2001. Concrete, reinforced and prestressed concrete structures Part 1; Design and construction. Berlin:
Beuth
Noakowski, P. & Schfer, H.G. 2003: Steifigkeitsorientierte
Statik im Stahlbetonbau. Berlin: Ernst & Sohn
Roetzer, J., Douglas, H. & Maurer, H. 2005 & 2006. Hazard
and Safety Investigations for LNG-Tanks. Part 1: Earthquakes. Part 2: Explosions. Part 3: Liquid Spill. LNGJournal